The subject of statistics

[Bungee Jumper]

Neither. 

 

You look at the Uninterested group’s response to the snack question, and throw out the Interested responses.  The Interested responses are biased, as they would have joined regardless of the snacks, as evidenced by their interest without the bribe (based on the wording of the question no snacks were offered).  You want to see whether the presence of snacks would have had any influence on the Uninterested respondents.  Your null hypothesis is that there is no influence, and any significant change from that would give you your answer.

817544[/snapback]

 

Of course, even that's flawed, given the way the questions were structured. No matter how you do it, you're comparing two different measurables: "interested/uninterested", and "1-5". So even in your case, you'd be measuring how "uninterested" turns into a numeric scale if snacks are offered...and what the hell kind of result would that give? "Everyone who was uninterested, when offered snack foods, would have had their opinion influenced by a measure of 2 on a 5-scale." Better to ask the second question "Would you have been interested or uninterested if snack foods were offered", so you can actually measure the effect the offer has (i.e. "23% of uninterested people were interested if snacks were available".)

 

And then, of course, there's the little fact that "intent" and "action" are two different things. If you're studying how offering snacks affects membership in a club, your measurable should be membership, not intent...

[Ramius]

Neither. 

 

You look at the Uninterested group’s response to the snack question, and throw out the Interested responses.  The Interested responses are biased, as they would have joined regardless of the snacks, as evidenced by their interest without the bribe (based on the wording of the question no snacks were offered).  You want to see whether the presence of snacks would have had any influence on the Uninterested respondents.  Your null hypothesis is that there is no influence, and any significant change from that would give you your answer.

817544[/snapback]

 

I saw this this morning, and had a similar line of thought johnny. From a scientific standpoint, both of the methods are wrong. (this isnt taking away from the simple fact that in this "experiment" they are working with opinions, not fact, and means that the experiment is utter bull sh-- to begin with.)

 

As you said with method 1, you need to toss out the "interested" responders, because they are going to go to the meeting regardless. That group accomplishes nothing. The interested group can be used as a control, but it is not used properly in method 1. It doesn't matter whether or not they prefer snacks.

 

With method 2, you are tossing out the entire experimental group. This whole experiment is wrong because it is focusing on the interested group, when it should be focusing on the interested group.

 

As i stated above, these types of experiments are complete bull sh--, because they are based solely off of opinion. There are way too many variables and controls needed to run an opinion experiment. Some psychology experiments actually involve measurable science, but i always had problems with opinion experiments.

 

However, if someone asked me to properly design this experiment, this is how it would be done.

 

Determining if snacks influence joining the chess club:

1. Get your group of random subjects.

2. Run a preliminary questionnaire asking whether or not the people in the group like snacks.

3. Anyone that doesnt like snacks is dismissed from the experiment. It's now established that everyone involved likes snacks.

4. Question everyone whether or not they are interested in chess.

5. Split the groups into interested and uninterested.

6. Further split the groups in half for each chess preference, ie- interested A, interested B, uninterested A, uninterested B.

7. Both group A's are invited to join the chess club, but no snacks are offered.

8. Both group B's are invited to join the chess club, and snacks ARE offered.

9. This experiment would need to be run multiple times with multiple groups of people.

 

Expected results: The interested groups act as your control. There should be no significant difference in chess club attendance for interested A vs interested B. If there is, your other results are not valid. The real results from the experiment lie in uninterested A vs uninterested B. If B has consistently significantly higher attendance over the course of many trials, then you can say that snacks affect chess club attendance.

 

Thats how to properly run the experiment. But then again, what do i know? i'm just a lowly masters level published scientist.

[Bungee Jumper]

I saw this this morning, and had a similar line of thought johnny. From a scientific standpoint, both of the methods are wrong. (this isnt taking away from the simple fact that in this "experiment" they are working with opinions, not fact, and means that the experiment is utter bull sh-- to begin with.)

 

As you said with method 1, you need to toss out the "interested" responders, because they are going to go to the meeting regardless. That group accomplishes nothing. The interested group can be used as a control, but it is not used properly in method 1.  It doesn't matter whether or not they prefer snacks.

 

With method 2, you are tossing out the entire experimental group. This whole experiment is wrong because it is focusing on the interested group, when it should be focusing on the interested group.

 

As i stated above, these types of experiments are complete bull sh--, because they are based solely off of opinion. There are way too many variables and controls needed to run an opinion experiment. Some psychology experiments actually involve measurable science, but i always had problems with opinion experiments.

 

However, if someone asked me to properly design this experiment, this is how it would be done.

 

Determining if snacks influence joining the chess club:

1. Get your group of random subjects.

2. Run a preliminary questionnaire asking whether or not the people in the group like snacks.

3. Anyone that doesnt like snacks is dismissed from the experiment. It's now established that everyone involved likes snacks.

4. Question everyone whether or not they are interested in chess.

5. Split the groups into interested and uninterested. 

6. Further split the groups in half for each chess preference, ie- interested A, interested B, uninterested A, uninterested B.

7. Both group A's are invited to join the chess club, but no snacks are offered.

8. Both group B's are invited to join the chess club, and snacks ARE offered.

9. This experiment would need to be run multiple times with multiple groups of people.

 

Expected results: The interested groups act as your control. There should be no significant difference in chess club attendance for interested A vs interested B. If there is, your other results are not valid. The real results from the experiment lie in uninterested A vs uninterested B. If B has consistently significantly higher attendance over the course of many trials, then you can say that snacks affect chess club attendance.

 

Thats how to properly run the experiment. But then again, what do i know? i'm just a lowly masters level published scientist.

817614[/snapback]

 

Ideally, too, you should run the same experiment for the dance club, to control for the possibility that chess players are overly influenced by snack foods...

[Ramius]

Ideally, too, you should run the same experiment for the dance club, to control for the possibility that chess players are overly influenced by snack foods... 

817628[/snapback]

 

very true. Mountain Dew is going to have a much higher impact on the chess club than the dance club or the outdoors club.

 

I'm interested to see what kind of (*^*&%^$^#response HA comes up with now, after coli, you, and me posted.

[Bungee Jumper]

very true. Mountain Dew is going to have a much higher impact on the chess club than the dance club or the outdoors club.

 

I'm interested to see what kind of (*^*&%^$^#response HA comes up with now, after coli, you, and me posted.

817635[/snapback]

 

I don't know why you'd even run such a study anyway. Just ask them what their parents' jobs are, and take all the ones whose dads are doctors, since they'll be the smartest.

[Tux of Borg]

Did they come because of the snacks?

817452[/snapback]

 

Yes, but you have to eat the snacks on the beach. You are not allowed to bring a cake on a plane. (as my roommate found out today)

[Bungee Jumper]

Yes, but you have to eat the snacks on the beach.  You are not allowed to bring a cake on a plane.  (as my roommate found out today)

817689[/snapback]

 

"I want this muther!@#$in' cake off my plane!" - Sam Jackson.

 

[Orton's Arm]

However, if someone asked me to properly design this experiment, this is how it would be done.

The experiment you've designed would produce solid results. However, the chess club that could afford such an expensive experiment would be rare indeed. In this example, the chess club was only able to afford a survey. It's looking at offering a number of different things (one of which is free snacks), but it can only afford to offer a few of the things it's considering. It wants to know which things it should offer to increase membership. While the chess club realizes the results won't be as reliable as they could have been with the experiment you describe, it at least wants you to use the survey data to tell it something.

 

Am I correct in concluding you'd use method 2, except that you'd throw out the Interested data instead of the Uninterested data?

[Just Jack]

Yes, but you have to eat the snacks on the beach.  You are not allowed to bring a cake on a plane.  (as my roommate found out today)

817689[/snapback]

What kind of cake? If it was a fruitcake, I see their point, some of those can be used as a weapon, they're so dense.

[Johnny Coli]

The experiment you've designed would produce solid results. However, the chess club that could afford such an expensive experiment would be rare indeed. In this example, the chess club was only able to afford a survey. It's looking at offering a number of different things (one of which is free snacks), but it can only afford to offer a few of the things it's considering. It wants to know which things it should offer to increase membership. While the chess club realizes the results won't be as reliable as they could have been with the experiment you describe, it at least wants you to use the survey data to tell it something.

 

Am I correct in concluding you'd use method 2, except that you'd throw out the Interested data instead of the Uninterested data?

817750[/snapback]

Look, man, far be it from me to rain on your "gotcha" thread, but you're attempting to use a flawed study as an example of how you know more about statistics than the other posters who called you out for using a flawed study to support your eugenics program in the other thread.

 

And therein lies the problem with both threads. You are trying to make your argument without fully understanding the concepts you are using to form your end of the debate, and that makes it very hard for someone who does know the concept(s) to try and argue with you. Your response to everything in the other thread was "show me the evidence." But it's like arguing with someone who just got hit on the head with a piece of cheese and then extrapolates that event into "the moon is made of cheese". You could try and tell them that there is overwhelming evidence to the contrary, but if they keep pointing to the wedge of cheese as all the evidence they need, then there's really no point in continuing the debate.

 

Stop pointing at the wedge of cheese.

[Ramius]

The experiment you've designed would produce solid results. However, the chess club that could afford such an expensive experiment would be rare indeed. In this example, the chess club was only able to afford a survey. It's looking at offering a number of different things (one of which is free snacks), but it can only afford to offer a few of the things it's considering. It wants to know which things it should offer to increase membership. While the chess club realizes the results won't be as reliable as they could have been with the experiment you describe, it at least wants you to use the survey data to tell it something.

 

Am I correct in concluding you'd use method 2, except that you'd throw out the Interested data instead of the Uninterested data?

817750[/snapback]

 

Nope, because both of the above described methods are flawed. Also, both methods are created based on assumptions of what 1 group will choose to do, which is one of the many reason why the experiment is bull sh--.

 

Method 1 will give you a high correlation between offering snacks, and joining the chess club, but as you have shown many times in the past, there is one fact you cannot seem to grasp.

 

CORRELATION DOES NOT EQUAL CAUSATION!!!

 

All you can say from method 1 is that there's a correlation. You dont know the causation, because you only polled the "interested group". They could be coming due to the snacks, they also could be coming due to their interest in chess.

 

I work in science. We deal with hypotheses. I figure out what i THINK will happen, and test it. If it works, i have results. if not, you test something else. What i dont do is make an assumption, and then base my experiment off treating the assumption like it is the truth. thats what you have done consistently throughout your assinine arguements.

[Ramius]

Holcombs arm pwned by statistics AGAIN!

[Orton's Arm]

Nope, because both of the above described methods are flawed. Also, both methods are created based on assumptions of what 1 group will choose to do, which is one of the many reason why the experiment is bull sh--.

Suppose the chess club brought you in at the beginning. All they could afford was a survey. They asked you to write the survey questions, and interpret their results. They want to know which things (such as free snacks) would be most helpful in increasing membership. How would you go about helping them answer this question?

[Taro T]

There are so many flaws with this it isn't even funny.

 

How does the chess club go about offering "free" snacks to people? Are they getting a subsidy from the university and are trying to figure out how best to spend it? Are they pulling some sort of "Revenge of the Nerds" scam on the jock frat to get their "snacks"?

 

Rather than simply asking if people are "interested" in the chess club or not, might it not be good to find out if any of the people can play chess or want to learn?

 

What you need to do to increase membership will likely be HIGHLY effected by who the prospective members are and what their interest and aptitudes in chess are. I'd much rather know that 1st and then worry about whether they'd prefer snacks, pocket protectors, or pornos.

[Orton's Arm]

Look, man, far be it from me to rain on your "gotcha" thread, but you're attempting to use a flawed study as an example of how you know more about statistics than the other posters who called you out for using a flawed study to support your eugenics program in the other thread.

Show me where I've made the claim that I know more about statistics than, for example, you know or Ramius knows.

And therein lies the problem with both threads.  You are trying to make your argument without fully understanding the concepts you are using to form your end of the debate, and that makes it very hard for someone who does know the concept(s) to try and argue with you. 

I haven't offered an analysis of the problem yet, so your criticism is premature. My intention in starting this thread was to have an intelligent conversation about statistics, in a less emotional atmosphere than a thread about eugenics would have created. Partly I saw this thread as an opportunity for people to prove they understand the subject on a deeper level than, for example, the statistics-oriented professor I mentioned earlier. I also saw it as a chance for people to get an accurate measure of how well I understand the topic, instead of making the type of assumptions I've sometimes seen.

[Bungee Jumper]

What i dont do is make an assumption, and then base my experiment off treating the assumption like it is the truth. thats what you have done consistently throughout your assinine arguements.

817879[/snapback]

 

Although I want to hasten to add that many scientists do just that, then "massage the data" from the experiment to fit the assumption. That horrific Weiss study HA linked to earlier is a great example.

 

But that doesn't mean it's science. At best, it's really really bad science.

[Bungee Jumper]

Show me where I've made the claim that I know more about statistics than, for example, you know or Ramius knows.

And therein lies the problem with both threads.  You are trying to make your argument without fully understanding the concepts you are using to form your end of the debate, and that makes it very hard for someone who does know the concept(s) to try and argue with you. 

I haven't offered an analysis of the problem yet, so your criticism is premature. My intention in starting this thread was to have an intelligent conversation about statistics, in a less emotional atmosphere than a thread about eugenics would have created. Partly I saw this thread as an opportunity for people to prove they understand the subject on a deeper level than, for example, the statistics-oriented professor I mentioned earlier. I also saw it as a chance for people to get an accurate measure of how well I understand the topic, instead of making the type of assumptions I've sometimes seen.

817916[/snapback]

 

This is the second time you've used this phrase, so I have to ask: what the !@#$ is a "statistics-oriented professor"?

[Chilly]

I haven't offered an analysis of the problem yet, so your criticism is premature. My intention in starting this thread was to have an intelligent conversation about statistics, in a less emotional atmosphere than a thread about eugenics would have created. Partly I saw this thread as an opportunity for people to prove they understand the subject on a deeper level than, for example, the statistics-oriented professor I mentioned earlier. I also saw it as a chance for people to get an accurate measure of how well I understand the topic, instead of making the type of assumptions I've sometimes seen.

 

This is the second time you've used this phrase, so I have to ask: what the !@#$ is a "statistics-oriented professor"? 

817930[/snapback]

 

One who finds statistics, rather then men, women, animals, or anything else, sexually attractive.

[Orton's Arm]

This is the second time you've used this phrase, so I have to ask: what the !@#$ is a "statistics-oriented professor"? 

817930[/snapback]

A professor in a subject which involves a lot of statistical analysis.

[Bungee Jumper]

A professor in a subject which involves a lot of statistical analysis...

817940[/snapback]

 

...of what? Genetics? Fluid dynamics? Nuclear physics? Botany? Economics? Golf?